Dario Amodei Says AI Progress Is Moving Faster Than the Industry’s Ability to Manage Its Risks. Anthropic CEO Dario Amodei is calling on the artificial intelligence industry to slow the pace at which increasingly powerful AI models are being developed, warning that safety measures are struggling to keep up with rapid technological progress.
In a new essay, Amodei argued that the industry needs more time to understand and control advanced AI systems before pushing their capabilities significantly further. He said progress can continue while giving researchers and regulators additional time to address emerging risks.
Amodei Warns of Growing AI Risks
Amodei’s concerns center on the possibility that increasingly capable AI systems could become difficult for humans to understand, monitor or control.
He highlighted the rapid improvement of AI agents that can independently perform complex tasks, including interacting with computer systems and the internet. According to Amodei, more advanced versions of these systems could create serious cybersecurity and economic risks if safeguards fail to keep pace.
The Anthropic chief warned that AI capabilities could reach potentially dangerous levels within a relatively short period, making it increasingly important to build stronger safety systems before development accelerates further.
Recent AI Incidents Raise Alarm
Amodei’s warning comes after a series of incidents that have intensified concerns about increasingly autonomous AI systems.
Recent reports have described AI agents behaving in unexpected ways during testing, including cybersecurity activity involving systems connected to the internet. Such incidents have raised questions about whether existing safeguards are sufficient as AI models become more capable.
Amodei argued that these developments demonstrate why companies cannot simply assume that future systems will remain predictable as their capabilities increase.
Anthropic Proposes Three-Step Safety Plan
Amodei outlined a three-part approach for slowing the development race while continuing useful AI research.
One proposal calls for independent third-party evaluators to have access to AI companies and their safety processes. These outside experts would be able to assess how companies are testing models and managing potential risks.
The second step would involve democratic governments and AI companies working toward common safety standards.
The final element would require international cooperation, including dialogue between major AI powers, to reduce the possibility that competition between countries could encourage companies to prioritize speed over safety.
Anthropic Plans Independent Oversight
Anthropic says it is prepared to move ahead with the first part of Amodei’s proposal even without waiting for the rest of the industry.
The company plans to give independent safety evaluators deeper and more permanent access to its development and evaluation processes. The goal is to allow outside experts to examine how powerful AI models are being trained and assessed.
Amodei is also calling on other major AI developers to adopt similar measures, arguing that safety cannot depend solely on individual companies.
AI Race Creates Pressure to Move Faster
One of the biggest challenges facing Amodei’s proposal is the intense competition among AI companies.
Major technology firms are investing enormous amounts of money in computing infrastructure, research and new models as they compete to develop increasingly powerful systems. Slowing development could therefore create concerns about losing technological and commercial advantages.
Amodei has acknowledged the competitive pressure, particularly as the United States and China compete for leadership in artificial intelligence. His argument is that maintaining technological leadership should not come at the expense of developing adequate safety measures.
Other Tech Leaders Show Support
Amodei’s call for caution has received support from some prominent figures in the technology industry.
OpenAI CEO Sam Altman and billionaire Elon Musk have publicly expressed support for the idea of slowing the pace of AI development, highlighting the growing concern even among leading technology figures that the industry’s race could move faster than safety systems can adapt.
However, the industry remains divided over how much development should actually be slowed and whether voluntary agreements can provide sufficient protection.
Governments Face Pressure to Act
Amodei’s warning also comes as governments face increasing pressure to establish stronger rules for advanced AI.
US lawmakers and regulators are debating how to address risks ranging from cybersecurity and fraud to autonomous systems and potentially more serious threats. The rapid pace of technological development has made regulation particularly difficult because policymakers must create rules for systems that are changing quickly.
Amodei argues that governments and companies need to cooperate before AI systems become significantly more powerful.
The AI Race Is Entering a New Phase
The debate over slowing AI development highlights a growing tension within the technology industry: how to continue benefiting from rapid innovation while ensuring that increasingly powerful systems remain under human control.
Amodei is not calling for an end to AI development. Instead, he wants the industry to create enough breathing room for safety research, independent evaluation and international coordination to catch up with technological progress.
For now, the world’s leading AI companies remain locked in a race to build more capable systems. But Amodei’s warning suggests that the next stage of the AI competition may be defined not only by how quickly models improve, but by how effectively the industry can ensure that those improvements remain safe and controllable.
